Integrative proteomic profiling of extracellular vesicle‐enriched preparations with machine learning identifies disease activity‐associated signatures linked to adverse pregnancy outcomes in systemic lupus erythematosus

Abstract Pregnancy in systemic lupus erythematosus (SLE) is strongly linked to disease flare‐ups and adverse pregnancy outcomes (APO), but effective molecular markers to supplement routine clinical evaluation are limited. In this study, we integrated proteomic profiling of plasma extracellular vesicle (EV)‐enriched preparations, antenatal clinical modeling and targeted ELISA validation to identify candidate biomarkers. Label‐free LC–MS/MS analysis of 49 plasma samples from pregnant women identified disease activity‐associated remodeling of the EV‐enriched proteome: active SLE was characterized by a chromatin‐immune protein profile, whereas stable SLE displayed a signature related to plasma homeostasis. Based on this discovery, AGT, FETUB, SPP2, and FCN3 were selected as a four‐protein biomarker panel to represent activity‐associated remodeling and APO‐related risk. Using 465 SLE pregnancy episodes, an L1‐penalized logistic regression model based on routine antenatal variables predicted APO with an AUROC of 0.773. In the EV cohort, combining predefined EV module scores with the recalibrated clinical model resulted in an AUROC of 0.849, compared with 0.809 for the clinical model alone. Targeted ELISA showed the expression trends of these four candidate EV‐enriched proteins. In conclusion, plasma EV‐enriched protein signatures serve as valuable molecular tools for assessing disease activity and evaluating pregnancy risks among women with SLE.